The best marketing prompts are not clever one-liners - they are structured briefs that tell the model what role to play, what inputs it has, and exactly what to hand back. Below are 10 marketing prompts covering strategy, landing pages, content planning, paid creative, email, SEO, messaging, webinars, launches, and attribution. Each one is written as a fill-in-the-blank template you can paste straight into ChatGPT, Claude, or Gemini. This set is pulled from the AI Central Library, where we test prompt structures against what our 300,000+ readers actually run in production.
The 10 best marketing prompts
1. Market strategy intelligence
Use this before you write anything. It forces the model to produce positioning evidence rather than adjectives.
Act as a senior market strategist. Analyze this market to extract JTBD, unmet needs, differentiation, and positioning risks. ICP: [industry, size, geo, budget, roles, maturity]. Offer: [product, pricing, model]. Competitors: [3-5 names + 1-line claims]. Constraints: non-generic, evidence-led, concise. Output: 1-paragraph snapshot; 6 unmet needs (ranked); 5-point differentiation map (value prop, proof, counter-claim); top risks + mitigation; 3 message hypotheses (persona, pain, promise, proof, CTA). Ask up to 3 critical questions if inputs are insufficient.
The last line is the part most people delete, and it is the most valuable. Letting the model ask before it answers is what separates a usable brief from confident filler.
2. Conversion landing page outline
This one produces a section map and wireframe notes, not just copy, so a designer can act on it directly.
Act as a conversion copywriter + UX lead. Create a landing page for [offer] targeting [ICP] to achieve [goal]. Inputs: benefits [list], differentiators [list], social proof [logos/testimonials], objections [list], compliance [list]. Output: section map (above-the-fold promise+CTA; credibility; objections; pricing; FAQ+trust); 3 headline + 3 subhead options; benefits as "feature to benefit to business impact"; 3 CTAs by intent; 8-10 FAQs; copy length per section; simple wireframe notes.
The feature-to-benefit-to-business-impact chain is the workhorse here. It stops the model writing benefit statements that never connect to a number a buyer cares about.
3. 90-day content plan
A full editorial calendar with a repurposing matrix, which is the part teams usually skip and then regret.
Act as an editorial director. Build a 90-day content plan for [product] and [ICP]. Inputs: themes [list], keywords [list], channels [list], cadence [e.g., 3/wk], resources [team/tools], repurposing [flow]. Output: weekly themes; per asset: title, persona+stage, thesis, primary keyword, angle, format, distribution, CTA, success metric; repurposing matrix (source to derivatives x channels); 5 contrarian topics; 5 data-led thought-leadership angles.
Ask for the contrarian topics separately, as this prompt does. Left to its own devices a model will give you ten variations of the same safe take.
4. Performance creative
Ad concepts structured as testable territories rather than a pile of one-off ideas.
Act as a performance creative director. Produce ad concepts for [platforms] to drive [objective] for [ICP]. Inputs: value prop [1 line], proof [stat/testimonial/demo], compliance [list], assets [list], budget [low/med/high]. Output: 5 creative territories (1-sentence thesis each); per territory: 3 hooks, 3 bodies (pain to promise to proof to CTA), 3 CTAs, suggested visuals/motion beats (0-3-6-10s for video), first lines/overlays per platform; testing matrix (hook x visual x CTA); learning agenda with hypotheses and metrics.
The testing matrix and learning agenda are what make this reusable. You are not asking for ads, you are asking for an experiment design that happens to include ads.
5. 5-email nurture sequence
For turning a new subscriber or a raw lead into someone ready to hear an offer.
Act as a lifecycle email strategist. Build a 5-email nurture sequence for [ICP] moving from [entry point] to [conversion goal]. Inputs: offer [desc], core objections [list], proof assets [case studies/data/testimonials], send cadence [e.g., day 0, 2, 5, 9, 14], compliance [list]. Output: sequence map (job of each email, one sentence); per email: subject line x3, preview text, body copy, single CTA, and the one objection it removes; segmentation and exit rules; re-engagement branch for non-openers; success metrics per email (open, click, reply, conversion).
One job and one CTA per email is the rule that matters. Sequences fail when every email tries to do everything.
6. SEO article brief
A brief a writer can execute without a follow-up call.
Act as an SEO strategist. Create a brief for keyword "[primary]" with secondaries [list] for [ICP]. Intent: [informational/commercial]. Competitors: [3 URLs]. Output: SERP gap analysis; H1-H3 outline + word guidance; FAQs (People Also Ask); internal links; schema; 3 originality angles (data, frameworks, case studies); draft intro + conclusion; on-page checklist; callouts for graphics/tables.
The originality angles line is doing the heavy lifting. Without it you get a brief that reconstructs the existing top three results, which is exactly what will not outrank them. If you want to go deeper on prompt structure itself, read the 26 principles of prompt engineering.
7. Messaging house
The single source of truth every other asset should inherit from.
Act as a brand and product marketing lead. Build a messaging house for [product] targeting [ICP]. Inputs: category [desc], differentiators [list], proof points [data/case studies], competitor claims [3-5], objections [list], tone [desc]. Output: one-line positioning statement; 3 message pillars, each with a claim, 2 supporting proof points, and a counter-objection; value prop by persona [list personas]; boilerplate at 25, 50, and 100 words; approved and banned language list; 5 headline patterns that can be reused across channels.
The banned language list is underrated. Naming what you will not say is the fastest way to stop three channels drifting into three different voices.
8. Webinar conversion kit
Everything around the webinar, which is where the conversions actually happen.
Act as a demand gen manager. Build a webinar plan to drive [goal] for [offer] to [ICP]. Inputs: topic [desc], speakers [names/roles], timing [time zones/dates], promo channels [list], reg tool [name], follow-up offer [desc]. Output: 3 titles + abstracts; run-of-show with timestamps; slide storyline (titles + bullets); interactions (polls/Q&A/demo); promo assets (3 emails, 5 social posts, 2 partner blurbs); reg page copy; post-webinar automation (segments, 3 emails, sales notes); KPIs + benchmarks.
Note that the post-webinar automation is segmented. Sending the same follow-up to attendees and no-shows is the most common way to waste a webinar list.
9. 4-week launch plan with experiments
A launch plan where every channel activity is framed as a test with a decision rule attached.
Act as a growth marketer. Create a 4-week launch for [product/feature] across [email, paid social, SEO, community, influencer, PR]. ICP: [personas]. Inputs: budgets/channel, assets, compliance, baselines. Output: weekly plan by channel; per experiment: hypothesis, audience, message/creative, metric, sample size, duration, decision rule; dependencies + timeline; risk/mitigation; sales enablement brief (talk track, one-pager bullets, objections).
Decision rule is the key field. Deciding in advance what result kills a channel is what stops a launch running four weeks of activity with no conclusions.
10. Analytics attribution framework
The least glamorous prompt on the list and probably the highest leverage.
Act as a marketing analytics lead. Design measurement for [business model] with funnel [stages], channels [list], conversions [define], data sources [CRM, ads, web, product]. Constraints: privacy, offline, sampling. Output: KPI hierarchy (north star, input, guardrails); attribution recommendation (model + rationale); channel measurement plan; UTM/event taxonomy; dashboard spec (dimensions/metrics/breakdowns/filters); experiment governance (power/MDE/thresholds); monthly insights cadence; 3 diagnostic analysis examples.
Asking for a model plus rationale rather than just a model is deliberate. Attribution choices are trade-offs, and you want the reasoning documented so you can revisit it when the data changes.
How to get better results from these prompts
Four things separate a good output from a wasted one:
Fill every bracket. A prompt with half its inputs missing produces generic output, and the model will not tell you it is guessing
Paste real context. Actual competitor pages, real testimonials, your existing metrics - the specificity of the input caps the quality of the output
Let the model ask questions. Add "ask up to 3 clarifying questions before answering" to any of these if it is not already there
Run them in sequence. Prompt 1 feeds prompt 7, which feeds prompts 2, 4, and 5. Running them in isolation is why teams end up with four different versions of their positioning
If you are new to structured prompting, our setup guide for ChatGPT, Claude, and Copilot covers the groundwork first.
Frequently Asked Questions
Do these marketing prompts work in ChatGPT, Claude, and Gemini?
Yes. They are structured as role, inputs, constraints, and output format, which every major model handles well. You may need to adjust length expectations, since context limits differ between tools.
Which marketing prompt should I run first?
Start with the market strategy intelligence prompt, then the messaging house. Those two produce the positioning and language that every other prompt on this list should inherit.
Why are these prompts so long?
Because length here is specification, not padding. Each bracket removes a guess the model would otherwise make, and the output format section is what makes the result usable without rewriting.
Can I use these prompts for a small business or solo marketing?
Yes. Reduce the scope inside the brackets rather than shortening the prompt - a 4-week launch across two channels uses the same structure as one across six.
How do I stop the output being generic?
Fill every bracket with real detail, paste in actual competitor copy and real proof points, and add a constraint line banning generic phrasing. Generic input is the single biggest cause of generic output.
